IRC channel logs

<Pellescours>youpi: i don't know how to use the netfs library, but I don't see any call to netfs_startup in console-client. Does it seems valid?

<youpi>I don't know

<Pellescours>i'm trying to improve hurd doc (hurd.texi), so I'm adding a section for lidnetfs. But as I don't really have the knowledge I'm reading the usage and internal.

<Pellescours>and I saw that case

<Pellescours>by doing a grep, it seems to be the only one that initialize betfs but does not startup it

<Pellescours>probably a bug then. I will try to add it and check.

<Pellescours>ah but console is not a program that you start as a translator. So you don't have the bootstrap port

<Pellescours>i see that the server is started manually. That's fine then.

<solid_black>hi all

<solid_black>Pellescours: yes, that's it, console-client (aka /bin/console) is not started as a translator, so it doesn't have the bootstrap port, instead it explicitly calls file_name_lookup () + netfs_init () + file_set_translator ()

<Pellescours>the fact that the console node id not translated as dir, can it be bex1use the code explixitely ignore netfs calls on root node ? https://github.com/etienne02/hurd/blob/master/console-client/trans.c#L80

<solid_black>Pellescours: did you see my messages about this issue yesterday?

<solid_black>it shows up as a file and not a directory because of an issue in glibc

<solid_black>the console is doing everything right, glibc is at fault

<Pellescours>solid_black: I'm not sure because the node is set with no_trans https://github.com/etienne02/hurd/blob/master/console-client/trans.c#L879

<solid_black>not sure why it uses O_NOTRANS there, i.e. I don't think we expect the underlying node to already have a translator set

<solid_black>but that doesn't matter

<solid_black>what matters is coreutils's ls(1) passes AT_NO_AUTOMOUNT when looking /dev/cons up, and glibc translates that to O_NOTRANS, and that causes ext2fs to return its empty file node, not console's root node

<Pellescours>ok, I see

<Pellescours>any news for alpine hurd?

<solid_black>not really

<azert>Pellescours: let’s assume that solid_black is trying to fix a not-a-bug. And glibc does expected behavior. Wouldn’t creating the node as a directory fix your issue?

<azert>Like inside the console client

<p4r4D0xum>anybody experianced issues installing binutils due to lzma error?

<p4r4D0xum>it seems something is wrong with dpkg-deb. --fix-broken install fails miserably

IRC channel logs

2024-03-13.log